mapcache_seed: try to kill still running threads after grace period

[marisn]

On rare occasions a seeding thread might be stuck (e.g. deadlocked). In a such case seeder process can not finish at all as it would wait for thread to complete forever.

This PR waits for the thread to finish ten times with increasing waiting time till 55 seconds in total have passed and then proceeds to terminate all remaining threads.

Questions to discuss:

• should there be a warning logged/printed?
• are 55 seconds of grace time enough?

[jmckenna]

@marisn can you re-base this again main ? (I had made several changes to the CI system, so hopefully this can run without issues...)

[marisn]

I had already forgotten about this PR :D Let me know what else is needed.

It is really hard to test this PR, as it gets hit only when something goes terribly wrong (e.g. file system failure).

[jmckenna]

Thanks @marisn. I think a warning printed out when the 10x threshold is hit would be useful too.

[marisn]

Thanks @marisn. I think a warning printed out when the 10x threshold is hit would be useful too.

Is it better now?

[jmckenna]

Yes thanks @marisn